<p>The space–air–ground–sea integrated network (SAGSIN) integrates diverse heterogeneous networks, which makes the pervasive communication and computing service of maritime users possible. In SAGSIN, communication nodes are scarce and computing resources are tight. Tasks of users can be offloaded to the network edge, such as at surface ship, buoy and drone. However, some traditional task offloading decisions become inapplicable as users move or the topology of networks changes. Therefore, in order to ensure the continued effectiveness of the task offloading strategy in dynamic ocean scenarios, we need to design an efficient offloading method to improve the overall network performance. To address this issue, we propose a resource-aware deep reinforcement learning method to optimize task offloading decisions. In this scheme, the available resources of mobile users are perceived in real time, and efficient computation offloading and task migration decisions are made based on the results. Simulation results validate the effectiveness of the proposed algorithm. In addition, the simulation results show that the proposed algorithm can effectively reduce the system delay and energy consumption as the number of tasks increases, and reduce task migration.</p>